Studying Privacy Aspects of Learned Knowledge Bases in the Context of Synthetic and Medical Data.

INTRODUCTION: Retrieving comprehensible rule-based knowledge from medical data by machine learning is a beneficial task, e.g., for automating the process of creating a decision support system. While this has recently been studied by means of exception-tolerant hierarchical knowledge bases (i.e., knowledge bases, where rule-based knowledge is represented on several levels of abstraction), privacy concerns have not been addressed extensively in this context yet. However, privacy plays an important role, especially for medical applications. METHODS: When parts of the original dataset can be restored from a learned knowledge base, there may be a practically and legally relevant risk of re-identification for individuals. In this paper, we study privacy issues of exception-tolerant hierarchical knowledge bases which are learned from data. We propose approaches for determining and eliminating privacy issues of the learned knowledge bases. RESULTS: We present results for synthetic as well as for real world datasets. CONCLUSION: The results show that our approach effectively prevents privacy breaches while only moderately decreasing the inference quality.

Comparison and Incorporation of Reasoning and Learning Approaches for Cancer Therapy Research.

Representing knowledge in a comprehensible and maintainable way and transparently providing inferences thereof are important issues, especially in the context of applications related to artificial intelligence in medicine. This becomes even more obvious if the knowledge is dynamically growing and changing and when machine learning techniques are being involved. In this paper, we present an approach for representing knowledge about cancer therapies collected over two decades at St.-Johannes-Hospital in Dortmund, Germany. The presented approach makes use of InteKRator, a toolbox that combines knowledge representation and machine learning techniques, including the possibility of explaining inferences. An extended use of InteKRator's reasoning system will be introduced for being able to provide the required inferences. The presented approach is general enough to be transferred to other data, as well as to other domains. The approach will be evaluated, e. g., regarding comprehensibility, accuracy and reasoning efficiency.

Automated Creation of Expert Systems with the InteKRator Toolbox.

Expert systems have a long tradition in both medical informatics and artificial intelligence research. Traditionally, such systems are created by implementing knowledge provided by experts in a system that can be queried for answers. To automatically generate such knowledge directly from data, the lightweight InteKRator toolbox will be introduced here, which combines knowledge representation and machine learning approaches. The learned knowledge is represented in the form of rules with exceptions that can be inspected and that are easily comprehensible. An inference module allows for the efficient answering of queries, while at the same time offering the possibility of providing explanations for the inference results. The learned knowledge can be revised manually or automatically with new evidence after learning.